www.nature.com/scientificreports
OPEN
received: 15 December 2016 accepted: 09 February 2017 Published: 16 March 2017
Somatic genome editing with CRISPR/Cas9 generates and corrects a metabolic disease Kelsey E. Jarrett1,2, Ciaran M. Lee3, Yi-Hsien Yeh1, Rachel H. Hsu1, Rajat Gupta1, Min Zhang1, Perla J. Rodriguez2,4, Chang Seok Lee1, Baiba K. Gillard4,5, Karl-Dimiter Bissig6, Henry J. Pownall4,5, James F. Martin1,7, Gang Bao3 & William R. Lagor1,2 Germline manipulation using CRISPR/Cas9 genome editing has dramatically accelerated the generation of new mouse models. Nonetheless, many metabolic disease models still depend upon laborious germline targeting, and are further complicated by the need to avoid developmental phenotypes. We sought to address these experimental limitations by generating somatic mutations in the adult liver using CRISPR/Cas9, as a new strategy to model metabolic disorders. As proof-of-principle, we targeted the low-density lipoprotein receptor (Ldlr), which when deleted, leads to severe hypercholesterolemia and atherosclerosis. Here we show that hepatic disruption of Ldlr with AAV-CRISPR results in severe hypercholesterolemia and atherosclerosis. We further demonstrate that co-disruption of Apob, whose germline loss is embryonically lethal, completely prevented disease through compensatory inhibition of hepatic LDL production. This new concept of metabolic disease modeling by somatic genome editing could be applied to many other systemic as well as liver-restricted disorders which are difficult to study by germline manipulation. Metabolic disease modeling is an essential component of biomedical research and a mandatory prerequisite for the treatment of human disease. Genetically engineered mice are the most frequently used model organism to study metabolic liver diseases, and have been invaluable in advancing our knowledge of many human disorders. However, many metabolic genes also play important roles in development, and their loss may result in embryonic lethality. The mouse genome informatics (MGI) database describes that at least 30% of all homozygous mutant lines created are embryonic or perinatal lethal1. Specific examples include methylmalonic aciduria (Mmachc)2, mevalonic aciduria (Mvk)3, lipoprotein lipase deficiency (Lpl)4, and hypobetalipoproteinemia (Apob)5, which underlie devastating human liver diseases. In some cases, hypomorphic alleles have been developed that recapitulate key aspects of a phenotype, such as ornithine transcarbamylase (Otc) deficiency, where the spf mutant has 10%, and the Spfash mutant 5%, of normal liver enzymatic activity6. Nevertheless, establishing a hypomorphic mouse model is difficult and often does not fully recapitulate the human disease. Several methods exist to study liver-expressed metabolic genes in adult mice. The cleanest of these are conditional mouse models, where the gene of interest is deleted by a liver-specific CRE recombinase transgene or on demand with viral delivery of CRE. Although robust, this technology is time-consuming and not well suited to testing multiple candidate genes, either alone or in combination. Loss-of-function studies can also be performed through the use of chemically modified antisense oligonucleotides (ASOs)7, locked nucleic acids8, siRNA nanoparticles9, and viral expression of shRNA10. These methods have advantages for achieving acute knockdown of target genes, but all have unique caveats, and leave varying degrees of residual expression and activity. Recent advances suggest that somatic genome editing could be used for genetic loss-of-function in adult animals. Clustered Regularly-Interspaced Short Palindromic Repeats/Cas9 (CRISPR/Cas9) is a bacterial immune system that has been adapted for genome editing in mammalian cells11,12. Cas9 is a programmable nuclease that 1
Department of Molecular Physiology and Biophysics, Baylor College of Medicine, Houston, TX 77030, USA. Integrative Molecular and Biomedical Sciences Graduate Program, Baylor College of Medicine, Houston, TX 77030, USA. 3Department of Bioengineering, Rice University, Houston, TX 77030, USA. 4Houston Methodist Research Institute, Houston, TX 77030, USA. 5Weill Cornell Medicine, Houston, TX 77030, USA. 6Center for Cell and Gene Therapy, Department of Molecular and Cellular Biology, Baylor College of Medicine, Houston, Texas 77030, USA. 7 Texas Heart Institute, Houston, TX 77030, USA. Correspondence and requests for materials should be addressed to W.R.L. (email: [email protected]) 2
Scientific Reports | 7:44624 | DOI: 10.1038/srep44624
1
www.nature.com/scientificreports/ generates double-stranded breaks (DSB) in DNA dictated by binding of a ~20 nucleotide recombinant “guide RNA” (gRNA) to the target site. DSB’s produced by Cas9 are most often repaired through the cell’s error-prone non-homologous end joining (NHEJ) pathway, resulting in small insertions or deletions (indels). The vast majority of indels shift the reading frame13, introducing a premature stop codon or resulting in nonsense mediated decay of the mRNA - effectively “knocking out” a gene. We and others have shown that delivery of CRISPR/Cas9 can effectively knockout liver-expressed genes through methods such as hydrodynamic tail vein injection14 and lipid nanoparticles15. Viral vectors offer the advantage of near complete liver transduction, eliminating potential redundancy from neighboring hepatocytes. Adenoviral vectors have been used to deliver CRISPR/Cas9 to disrupt Pcsk916 in mouse liver. Adeno-Associated Viral (AAV) vectors have also been used for somatic genome editing in adult mouse liver17,18, including the identification of a smaller Cas9 ortholog from Staphylococcus aureus (SaCas9) that can edit genes with high specificity18. Collectively, these studies suggest that CRISPR/Cas9 genome editing is very efficient in murine hepatocytes. Somatic genome editing using CRISPR/Cas9 might be used to establish novel metabolic disease models. Hypercholesterolemia and associated atherosclerosis is a major problem in the developed world, and would serve as an excellent test case for modeling a common metabolic disease. The liver secretes triglyceride-rich very low density lipoprotein (VLDL) particles into the circulation to provide energy to the rest of the body. These particles each contain one Apolipoprotein B (APOB) molecule as their core structural component, an essential player in export of excess triglyceride from the liver. VLDL particles are broken down by lipases in the circulation, generating smaller, cholesterol-enriched, low density lipoprotein (LDL) particles. Excess LDL is normally removed through the action of the low density lipoprotein receptor (LDLR), which is highly expressed in the liver, where it is believed to account for approximately 70% of LDL clearance19,20. Mutations in the LDLR gene are the basis of Familial Hypercholesterolemia (FH) (MIM 143890), a Mendelian lipid disorder characterized by exceptionally high levels of LDL cholesterol (LDL-C), xanthomas, and severe atherosclerotic vascular disease. We previously showed that human FH could be modeled in chimeric mice through transplantation with hepatocytes from an FH patient, and the ensuing hypercholesterolemia corrected by additive gene therapy with AAV9 vectors21. Here we generate the first model of metabolic liver disease via somatic genome editing. Mice treated with AAV-CRISPR vectors to somatically disrupt the Ldlr gene in the liver develop a severe hypercholesterolemia as well as a systemic metabolic phenotype- atherosclerosis. Concomitant disruption of Apob is protective from hypercholesterolemia and atherosclerosis, and demonstrates the utility of somatic genome editing to validate potential therapeutic targets for metabolic liver disorders.
Results
Somatic disruption of metabolic genes with AAV-CRISPR. Small guide RNAs were designed using
E-CRISP (http://www.e-crisp.org) targeting either exon 14 of Ldlr or exon 5 of Apob. The gRNAs were cloned into our AAV vector that expresses the gRNA from the U6 promoter, upstream of green fluorescent reporter driven by the ubiquitous chicken beta (CB) actin promoter. Male Cas9 transgenic mice22 (6–9 weeks old) were injected with 6 × 1011 genome copies (GC) of AAV8 encoding the different gRNA’s. Mice were randomly assigned one of three groups: (i) BbsI (empty cloning site as a “nontargeting” control, 6 × 1011 GC), (ii) Ldlr gRNA + BbsI (3 × 1011 each), and (iii) Ldlr gRNA + Apob gRNA (3 × 1011 each). The animals were placed on Western diet, a high fat cholesterol-rich diet, followed for 20 weeks, and sacrificed to assess atherosclerosis, plasma cholesterol, Ldlr and Apo B protein levels, and editing efficiency (Fig. 1A). Mice in groups receiving a gRNA to Ldlr had greater than 50% decreases in Ldlr mRNA expression (Fig. 1B). Animals receiving the Apob gRNA showed more than 90% knockdown of Apob mRNA (Fig. 1B). For both the Ldlr and Apob transcripts, similar results were obtained using primers upstream and downstream of the cut site. LDLR protein was readily detectable in livers from the BbsI control group, and nearly completely absent from the Ldlr gRNA and Ldlr + Apob gRNA groups (Fig. 1C). The Ldlr gRNA group had a dramatic increase in Apo B-100 and Apo B-48 compared to the control animals, while the Ldlr + Apob gRNA treated mice had a nearly complete loss of Apo B protein (Fig. 1D).
Efficiency and specificity of somatic genome editing. The efficiency and specificity of editing was
assessed by deep sequencing of livers following the 20 week study. In addition to the target regions of Ldlr and Apob, we used the COSMID algorithm23 to predict potential off-target sites in the mouse genome. On- and off-target sites were amplified by PCR to generate barcoded libraries (n = 4 per group) (Supplementary Table 1). Deep sequencing revealed a high degree of editing in the Ldlr gene in both groups receiving the Ldlr gRNA with a mean indel formation of 54.3 ± 16.6% (Fig. 2A). In the case of Apob, the group receiving a gRNA targeting this gene had an impressive rate of indel formation of 74.1 ± 23.4% (Fig. 2A). The most common mutations were small insertions and deletions occurring around positions −3 to −5 relative to the Protospacer Adjacent Motif (PAM) (Fig. 2B). One of the predicted off-target sites for the Ldlr gene (OT 11) had a low frequency of mutations which was detectable above background at 5.12 ± 2.02% (Fig. 2C). This particular off-target site is within an intron of the Stx8 gene. No detectable off-target editing was detected for any predicted sites for the Apob gRNA (Fig. 2D). However, indel formation was observed at several highly similar predicted sites in mice that received the BbsI stuffer sequence present in many CRISPR plasmids24, ranging from 1.3–30% depending on location (Supplementary Table 2). Perhaps most interestingly, small insertion events containing sequences from the AAV inverted terminal repeats (ITRs) were detected at the cut site with the highest editing efficiency (exon 5 of the Apob gene), accounting for 10–26% of the total reads (Fig. 3).
Metabolic disease modeling and correction with CRISPR/Cas9. Mice receiving the control virus displayed a gradual rise in plasma cholesterol from the Western diet, reaching a maximum of 350 ± 18.7 mg/dL after 20 weeks. The Ldlr gRNA group had significantly higher plasma cholesterol which increased to a maximum of 728 ± 174 mg/dL. In contrast, mice receiving both the Ldlr and Apob gRNAs exhibited a rapid drop in plasma Scientific Reports | 7:44624 | DOI: 10.1038/srep44624
2
www.nature.com/scientificreports/
Figure 1. Somatic disruption of Ldlr and Apob. (A) Experimental design- Male C57BL/6 J mice were injected with a total of 6 × 1011 genome copies of AAV, placed on Western Diet, and followed for 20 weeks. (B) Expression of mRNA Ldlr and Apob in the liver using primers upstream and downstream of the gRNA site. (C) LDLR and beta tubulin Western blot in mouse liver. (D) Apo B protein levels in plasma. Uncropped Western blot images are shown in Supplementary Data. For all panels: BbsI control n = 5, Ldlr gRNA + BbsI n = 5, Ldlr + Apob gRNA n = 6; data are represented as mean +/−S.D. and *p
OPEN
received: 15 December 2016 accepted: 09 February 2017 Published: 16 March 2017
Somatic genome editing with CRISPR/Cas9 generates and corrects a metabolic disease Kelsey E. Jarrett1,2, Ciaran M. Lee3, Yi-Hsien Yeh1, Rachel H. Hsu1, Rajat Gupta1, Min Zhang1, Perla J. Rodriguez2,4, Chang Seok Lee1, Baiba K. Gillard4,5, Karl-Dimiter Bissig6, Henry J. Pownall4,5, James F. Martin1,7, Gang Bao3 & William R. Lagor1,2 Germline manipulation using CRISPR/Cas9 genome editing has dramatically accelerated the generation of new mouse models. Nonetheless, many metabolic disease models still depend upon laborious germline targeting, and are further complicated by the need to avoid developmental phenotypes. We sought to address these experimental limitations by generating somatic mutations in the adult liver using CRISPR/Cas9, as a new strategy to model metabolic disorders. As proof-of-principle, we targeted the low-density lipoprotein receptor (Ldlr), which when deleted, leads to severe hypercholesterolemia and atherosclerosis. Here we show that hepatic disruption of Ldlr with AAV-CRISPR results in severe hypercholesterolemia and atherosclerosis. We further demonstrate that co-disruption of Apob, whose germline loss is embryonically lethal, completely prevented disease through compensatory inhibition of hepatic LDL production. This new concept of metabolic disease modeling by somatic genome editing could be applied to many other systemic as well as liver-restricted disorders which are difficult to study by germline manipulation. Metabolic disease modeling is an essential component of biomedical research and a mandatory prerequisite for the treatment of human disease. Genetically engineered mice are the most frequently used model organism to study metabolic liver diseases, and have been invaluable in advancing our knowledge of many human disorders. However, many metabolic genes also play important roles in development, and their loss may result in embryonic lethality. The mouse genome informatics (MGI) database describes that at least 30% of all homozygous mutant lines created are embryonic or perinatal lethal1. Specific examples include methylmalonic aciduria (Mmachc)2, mevalonic aciduria (Mvk)3, lipoprotein lipase deficiency (Lpl)4, and hypobetalipoproteinemia (Apob)5, which underlie devastating human liver diseases. In some cases, hypomorphic alleles have been developed that recapitulate key aspects of a phenotype, such as ornithine transcarbamylase (Otc) deficiency, where the spf mutant has 10%, and the Spfash mutant 5%, of normal liver enzymatic activity6. Nevertheless, establishing a hypomorphic mouse model is difficult and often does not fully recapitulate the human disease. Several methods exist to study liver-expressed metabolic genes in adult mice. The cleanest of these are conditional mouse models, where the gene of interest is deleted by a liver-specific CRE recombinase transgene or on demand with viral delivery of CRE. Although robust, this technology is time-consuming and not well suited to testing multiple candidate genes, either alone or in combination. Loss-of-function studies can also be performed through the use of chemically modified antisense oligonucleotides (ASOs)7, locked nucleic acids8, siRNA nanoparticles9, and viral expression of shRNA10. These methods have advantages for achieving acute knockdown of target genes, but all have unique caveats, and leave varying degrees of residual expression and activity. Recent advances suggest that somatic genome editing could be used for genetic loss-of-function in adult animals. Clustered Regularly-Interspaced Short Palindromic Repeats/Cas9 (CRISPR/Cas9) is a bacterial immune system that has been adapted for genome editing in mammalian cells11,12. Cas9 is a programmable nuclease that 1
Department of Molecular Physiology and Biophysics, Baylor College of Medicine, Houston, TX 77030, USA. Integrative Molecular and Biomedical Sciences Graduate Program, Baylor College of Medicine, Houston, TX 77030, USA. 3Department of Bioengineering, Rice University, Houston, TX 77030, USA. 4Houston Methodist Research Institute, Houston, TX 77030, USA. 5Weill Cornell Medicine, Houston, TX 77030, USA. 6Center for Cell and Gene Therapy, Department of Molecular and Cellular Biology, Baylor College of Medicine, Houston, Texas 77030, USA. 7 Texas Heart Institute, Houston, TX 77030, USA. Correspondence and requests for materials should be addressed to W.R.L. (email: [email protected]) 2
Scientific Reports | 7:44624 | DOI: 10.1038/srep44624
1
www.nature.com/scientificreports/ generates double-stranded breaks (DSB) in DNA dictated by binding of a ~20 nucleotide recombinant “guide RNA” (gRNA) to the target site. DSB’s produced by Cas9 are most often repaired through the cell’s error-prone non-homologous end joining (NHEJ) pathway, resulting in small insertions or deletions (indels). The vast majority of indels shift the reading frame13, introducing a premature stop codon or resulting in nonsense mediated decay of the mRNA - effectively “knocking out” a gene. We and others have shown that delivery of CRISPR/Cas9 can effectively knockout liver-expressed genes through methods such as hydrodynamic tail vein injection14 and lipid nanoparticles15. Viral vectors offer the advantage of near complete liver transduction, eliminating potential redundancy from neighboring hepatocytes. Adenoviral vectors have been used to deliver CRISPR/Cas9 to disrupt Pcsk916 in mouse liver. Adeno-Associated Viral (AAV) vectors have also been used for somatic genome editing in adult mouse liver17,18, including the identification of a smaller Cas9 ortholog from Staphylococcus aureus (SaCas9) that can edit genes with high specificity18. Collectively, these studies suggest that CRISPR/Cas9 genome editing is very efficient in murine hepatocytes. Somatic genome editing using CRISPR/Cas9 might be used to establish novel metabolic disease models. Hypercholesterolemia and associated atherosclerosis is a major problem in the developed world, and would serve as an excellent test case for modeling a common metabolic disease. The liver secretes triglyceride-rich very low density lipoprotein (VLDL) particles into the circulation to provide energy to the rest of the body. These particles each contain one Apolipoprotein B (APOB) molecule as their core structural component, an essential player in export of excess triglyceride from the liver. VLDL particles are broken down by lipases in the circulation, generating smaller, cholesterol-enriched, low density lipoprotein (LDL) particles. Excess LDL is normally removed through the action of the low density lipoprotein receptor (LDLR), which is highly expressed in the liver, where it is believed to account for approximately 70% of LDL clearance19,20. Mutations in the LDLR gene are the basis of Familial Hypercholesterolemia (FH) (MIM 143890), a Mendelian lipid disorder characterized by exceptionally high levels of LDL cholesterol (LDL-C), xanthomas, and severe atherosclerotic vascular disease. We previously showed that human FH could be modeled in chimeric mice through transplantation with hepatocytes from an FH patient, and the ensuing hypercholesterolemia corrected by additive gene therapy with AAV9 vectors21. Here we generate the first model of metabolic liver disease via somatic genome editing. Mice treated with AAV-CRISPR vectors to somatically disrupt the Ldlr gene in the liver develop a severe hypercholesterolemia as well as a systemic metabolic phenotype- atherosclerosis. Concomitant disruption of Apob is protective from hypercholesterolemia and atherosclerosis, and demonstrates the utility of somatic genome editing to validate potential therapeutic targets for metabolic liver disorders.
Results
Somatic disruption of metabolic genes with AAV-CRISPR. Small guide RNAs were designed using
E-CRISP (http://www.e-crisp.org) targeting either exon 14 of Ldlr or exon 5 of Apob. The gRNAs were cloned into our AAV vector that expresses the gRNA from the U6 promoter, upstream of green fluorescent reporter driven by the ubiquitous chicken beta (CB) actin promoter. Male Cas9 transgenic mice22 (6–9 weeks old) were injected with 6 × 1011 genome copies (GC) of AAV8 encoding the different gRNA’s. Mice were randomly assigned one of three groups: (i) BbsI (empty cloning site as a “nontargeting” control, 6 × 1011 GC), (ii) Ldlr gRNA + BbsI (3 × 1011 each), and (iii) Ldlr gRNA + Apob gRNA (3 × 1011 each). The animals were placed on Western diet, a high fat cholesterol-rich diet, followed for 20 weeks, and sacrificed to assess atherosclerosis, plasma cholesterol, Ldlr and Apo B protein levels, and editing efficiency (Fig. 1A). Mice in groups receiving a gRNA to Ldlr had greater than 50% decreases in Ldlr mRNA expression (Fig. 1B). Animals receiving the Apob gRNA showed more than 90% knockdown of Apob mRNA (Fig. 1B). For both the Ldlr and Apob transcripts, similar results were obtained using primers upstream and downstream of the cut site. LDLR protein was readily detectable in livers from the BbsI control group, and nearly completely absent from the Ldlr gRNA and Ldlr + Apob gRNA groups (Fig. 1C). The Ldlr gRNA group had a dramatic increase in Apo B-100 and Apo B-48 compared to the control animals, while the Ldlr + Apob gRNA treated mice had a nearly complete loss of Apo B protein (Fig. 1D).
Efficiency and specificity of somatic genome editing. The efficiency and specificity of editing was
assessed by deep sequencing of livers following the 20 week study. In addition to the target regions of Ldlr and Apob, we used the COSMID algorithm23 to predict potential off-target sites in the mouse genome. On- and off-target sites were amplified by PCR to generate barcoded libraries (n = 4 per group) (Supplementary Table 1). Deep sequencing revealed a high degree of editing in the Ldlr gene in both groups receiving the Ldlr gRNA with a mean indel formation of 54.3 ± 16.6% (Fig. 2A). In the case of Apob, the group receiving a gRNA targeting this gene had an impressive rate of indel formation of 74.1 ± 23.4% (Fig. 2A). The most common mutations were small insertions and deletions occurring around positions −3 to −5 relative to the Protospacer Adjacent Motif (PAM) (Fig. 2B). One of the predicted off-target sites for the Ldlr gene (OT 11) had a low frequency of mutations which was detectable above background at 5.12 ± 2.02% (Fig. 2C). This particular off-target site is within an intron of the Stx8 gene. No detectable off-target editing was detected for any predicted sites for the Apob gRNA (Fig. 2D). However, indel formation was observed at several highly similar predicted sites in mice that received the BbsI stuffer sequence present in many CRISPR plasmids24, ranging from 1.3–30% depending on location (Supplementary Table 2). Perhaps most interestingly, small insertion events containing sequences from the AAV inverted terminal repeats (ITRs) were detected at the cut site with the highest editing efficiency (exon 5 of the Apob gene), accounting for 10–26% of the total reads (Fig. 3).
Metabolic disease modeling and correction with CRISPR/Cas9. Mice receiving the control virus displayed a gradual rise in plasma cholesterol from the Western diet, reaching a maximum of 350 ± 18.7 mg/dL after 20 weeks. The Ldlr gRNA group had significantly higher plasma cholesterol which increased to a maximum of 728 ± 174 mg/dL. In contrast, mice receiving both the Ldlr and Apob gRNAs exhibited a rapid drop in plasma Scientific Reports | 7:44624 | DOI: 10.1038/srep44624
2
www.nature.com/scientificreports/
Figure 1. Somatic disruption of Ldlr and Apob. (A) Experimental design- Male C57BL/6 J mice were injected with a total of 6 × 1011 genome copies of AAV, placed on Western Diet, and followed for 20 weeks. (B) Expression of mRNA Ldlr and Apob in the liver using primers upstream and downstream of the gRNA site. (C) LDLR and beta tubulin Western blot in mouse liver. (D) Apo B protein levels in plasma. Uncropped Western blot images are shown in Supplementary Data. For all panels: BbsI control n = 5, Ldlr gRNA + BbsI n = 5, Ldlr + Apob gRNA n = 6; data are represented as mean +/−S.D. and *p
